AI Model Reviewer

Harness lane

Same model, same prompt, different harness: the vendor's own, a neutral one, and the raw API.

Each card shows a random output from that model; reload for another.

Four-stroke engine cutaway: crank, piston and cam-driven valves (Three.js)
Four-stroke engine cutaway: crank, piston and cam-driven valves (Three.js)

Claude Opus 5.5

Anthropic

283 finished runs ·

Four-stroke engine cutaway: crank, piston and cam-driven valves (Three.js)
Four-stroke engine cutaway: crank, piston and cam-driven valves (Three.js)

GPT-6 Astra

OpenAI

260 finished runs ·

Algoviz balanced trees
Algoviz balanced trees

DeepSeek V4.1 Flash

DeepSeek

219 finished runs ·

Algoviz balanced trees
Algoviz balanced trees

GPT-6 Sol

OpenAI

238 finished runs ·

Four-stroke engine cutaway: crank, piston and cam-driven valves (Three.js)
Four-stroke engine cutaway: crank, piston and cam-driven valves (Three.js)

Claude Fable 5.1

Anthropic

161 finished runs ·

Threejs mobile touch arcade game
Threejs mobile touch arcade game

Kimi K3

Moonshot

95 finished runs ·

Hot air balloon festival over rolling fields (Three.js)
Hot air balloon festival over rolling fields (Three.js)

Claude Sonnet 5.5

Anthropic

54 finished runs ·

Four-stroke engine cutaway: crank, piston and cam-driven valves (Three.js)
Four-stroke engine cutaway: crank, piston and cam-driven valves (Three.js)

gpt-6.1-sol

OpenAI

29 finished runs ·

Dolphin vlog bicycle
Dolphin vlog bicycle

Muse Spark 1.3

Meta

23 finished runs ·

Dolphin vlog bicycle
Dolphin vlog bicycle

Space Bunny Alpha

Undisclosed (stealth)

23 finished runs ·

Isometric SVG town with sprite sheet
Isometric SVG town with sprite sheet

Gemini 3.8 Flash

Google

8 finished runs ·

Pelican riding a bicycle (control)
Pelican riding a bicycle (control)

Grok 4.7

xAI

10 finished runs ·

Game battleship
Game battleship

Claude Haiku 5.5

Anthropic

7 finished runs ·